JP6283772B2 - Method and system for verifying mark reliability - Google Patents

Description

Cross-reference of related applications
[0001] This application is a continuation-in-part of US patent application Ser. No. 14 / 561,215 filed on Dec. 4, 2014. This application also claims priority to US Provisional Patent Application No. 62/053905, filed September 23, 2014.

[0002] The present disclosure relates generally to machine vision technology, and more particularly to a method and system for verifying the reliability of a mark.

[0003] Unfortunately, counterfeit products are widely available and are often difficult to distinguish. When counterfeiters produce fake merchandise, they typically copy labels that may include barcodes or other types of marks in addition to the actual product. On the surface, counterfeit marks may appear genuine and may even yield valid data when scanned (eg, deciphered to be an appropriate universal product code). Many of the currently available techniques for dealing with such copies rely on the basic idea of visually comparing an image of a mark that may be counterfeited with an image of the original mark.

[0004] When the product on which the mark is located is processed, packed, shipped, or the like, the mark may be mechanically damaged. Damage can occur in some unknown manner, resulting in unpredictable changes to the mark. The usual forms of damage include scratches, ink smears, ink that has been removed from the surface, and stains that blur some of the mark, any of which is when the mark is genuine. However, there is a possibility of deforming the mark so that it is no longer similar to the original. As a result, the process of performing the comparison (eg, image comparison software) may mistakenly consider the mark being examined as counterfeit (“false negative”).

[0005] Although the appended claims detail the features of the present technique, the technique can be best understood from the following detailed description in conjunction with the accompanying drawings.

[0006] An example of a mark. [0007] FIG. 2 is a view of the mark in FIG. 1 where the edge features of the mark have been extracted for clarity. [0008] FIG. 2 is a diagram of an example that may be a second forgery of the mark of FIG. [0009] FIG. 4 is a view of the mark in FIG. 3 where the edge features of the mark have been extracted for clarity. [0010] Figure 2 is an example of a two-dimensional data matrix printed by a heat transfer process. [0011] FIG. 5 compares features in the upper left portion of FIGS. 2 and 4; [0012] FIG. 2 is a block diagram of a computer system. [0013] FIG. 1 is a block diagram of a computer system according to one embodiment. [0014] FIG. 6 is a flowchart illustrating operations performed according to one embodiment. [0015] FIG. 6 is a flowchart illustrating operations performed in accordance with one embodiment. [0016] FIG. 3 is a one-dimensional barcode that illustrates several features that can be used in one embodiment. [0017] FIG. 6 is a graph of a polynomial approximation of an autocorrelation sequence for an original mark with candidate marks (evaluating authentic), according to one embodiment. [0018] FIG. 13 is a power series chart for the original mark data of FIG. [0019] FIG. 13 is a power series chart for the candidate mark data of FIG. [0020] FIG. 6 is a graph of a polynomial approximation of an autocorrelation sequence for an original mark with candidate marks (evaluating forgery), according to one embodiment. [0021] FIG. 16 is a power series chart for the candidate mark data of FIG. [0022] An example of an undamaged genuine candidate mark having the form of a printed two-dimensional barcode. [0023] FIG. 18 is a first example of the mark of FIG. 17 after being damaged. [0024] FIG. 18 is a second example of the mark of FIG. 17 after being damaged. [0025] FIG. 18 is a plot of the correlation of the values of the metric series representing the candidate mark signature of FIG. 17 against the stored series of metrics representing the original mark signature showing a match. [0026] Correlation of the value of the metric sequence representing the signature of the candidate mark in FIG. 18 to the stored sequence of metrics representing the signature for the genuine mark, indicating no match (false negative in this example) It is a plot. [0027] The correlation of the values of the metric sequence representing the signature of the candidate mark of FIG. 19 to the stored sequence of metrics representing the signature for the genuine mark, indicating no match (false negative in this example) It is a plot. [0028] For a stored series of metrics representing signatures for genuine marks (after removing metrics corresponding to removed candidate signature metrics) that showed a match (metrics indicating damage were removed FIG. 19 is a plot of correlation of values of a series of metrics representing the signature of the candidate mark of FIG. 18 (later). [0029] For a stored sequence of metrics that represent a signature for a genuine mark (after removing the metric corresponding to the removed candidate signature metric) indicating a match (in this example, the correct result), FIG. 20 is a plot of correlation of values of a metric series representing the signature of the candidate mark of FIG. 19 (after removing a metric indicating damage). [0030] FIG. 25 shows a match (in this example, the correct result) for a stored sequence of metrics that represent a signature for a genuine mark (after removing the metric corresponding to the removed metric) FIG. 21 is a plot of the correlation of values of a series of metrics representing the signature of the candidate mark of FIG. 20 (after removing a metric indicating damage). [0031] FIG. 6 is a flowchart illustrating operations performed in accordance with one embodiment. [0032] FIG. 6 illustrates an example of a set of data metrics according to one embodiment. [0033] FIG. 6 is a flowchart illustrating operations performed in accordance with each embodiment. It is a flowchart which shows the operation | movement performed according to each embodiment. [0034] FIG. 6 is a block diagram of a hardware environment according to one embodiment. [0035] FIG. 6 is a block diagram of a computer system according to one embodiment. [0036] FIG. 3 is a block diagram of a storage system according to one embodiment.

[0037] The present disclosure is generally directed to a method for verifying the reliability of a mark. In one embodiment, the processor (1) receives an image of a candidate mark from an image acquisition device and (2) uses the image to measure one or more characteristics at multiple locations in the candidate mark, resulting in Obtaining a first set of metrics; (3) removing a metric having a dominant amplitude from the first set of metrics, resulting in a trimmed first set of metrics; and (4) a computer readable memory. From which a second set of metrics representing one or more characteristics measured at a plurality of locations on the original mark is removed and (5) removed from the first set of metrics from the second set of metrics Metric corresponding to the selected metric, resulting in a second trimmed set of metrics, Comparing the first set that is timing, and a second set trimmed metrics (7) based on the comparison, candidate mark determines whether genuine.

[0038] According to another embodiment, the processor compares the first set of metrics to the second set of metrics and determines whether the candidate mark is genuine based on the comparison. , If it is determined that the candidate mark is not authentic, the processor performs the following additional steps: (1) removing the metric with the dominant amplitude from the first set, resulting in a metric Obtaining a trimmed first set; (2) removing a corresponding metric from the second set, resulting in a trimmed second set of metrics; (3) comparing the metric trimmed first set with the metric trimmed second set; and (4) whether the candidate mark is genuine based on the comparison. And a step of determining.

[0039] According to various embodiments, instead of sorting the signature data based on the saliency of each artifact in the original learned mark, the processor (1) saliency of each artifact in the signature data of the candidate mark. Are used as masters to compare the signature data of the original learned marks, (2) sort both sets of signature data according to their ordering, and (3) start with the most prominent artifacts, The signature data is gradually trimmed downward, and (4) a statistical correlation is performed between the signature data of the original mark and the signature data of the candidate mark. In one embodiment, trimming the data means reducing the correlated data by removing the second largest magnitude artifact from each successive statistical correlation. In other words, in one embodiment, the trimming of data removes the metric having the largest magnitude from the metric set of candidate mark signature data and the corresponding metric from the metric set of original mark signature data. Also including removing. Since the area of mechanical damage tends to be among the largest magnitude features in the signature data set, this removes problematic corruption from the signature data and lower true signature data Has the effect of allowing the underlying correlation to appear.

[0040] Some marks to which the techniques described herein can be applied include two-dimensional barcodes, such as QR codes, examples of which are shown in FIGS. Some two-dimensional barcodes include built-in error correction or redundancy. In a two-dimensional code or other mark that incorporates Reed-Solomon error correction or other redundancy, the removal of damaged signals in various embodiments is by utilizing information provided by error correction or other redundancy. Made easy. Such code identifies the damaged area as part of the function that corrects missing code information, so that information can be used to identify it as damaged before performing the signature analysis correlation step. Can be directly excluded.

[0041] According to another embodiment, the processor performs successive divisions or successive subdivisions on the metric of the candidate mark to remove the damage signal. In one example where the mark has a straight edge to be analyzed, such as a two-dimensional bar code finder bar, the processor places the damaged finder bar in two compartments, then four compartments, then the next. Subdivide continuously into 8 sections. The processor observes these individual signature correlations while continuously subdividing the partitions and sub-partitions.

[0042] The present disclosure often refers to "marks". As used herein, a “mark” is visible in some parts of the electromagnetic spectrum, but is not necessarily visible to the naked eye and is intentionally placed on a physical object. The mark can be used to uniquely identify the item, such as a serial number, or it can be used for some other purpose such as branding (eg, logo), labeling, or decoration. The mark can be printed, etched, shaped, formed, transferred, or otherwise affixed to the item using a variety of processes. The mark can be obtained so that it can be processed in electronic form. Possible devices that can be used for electronic acquisition of marks include machine vision cameras, bar code readers, line scan imagers, flatbed scanners, and handheld portable imaging devices. The mark is a barcode (eg, an international organization for standardization (“ISO”) and a two-dimensional (“2D”) barcode as specified in the International Electrotechnical Commission (“IEC”) standard ISO / IEC16022). , Tracking information such as expiration date or serial number).

[0043] In various embodiments, the original authentic mark includes artifacts that can be characterized by data often referred to herein as "signals", "signatures" or "signature signals". In addition, the candidate mark includes artifacts that can similarly represent characteristics by data, regardless of whether the candidate mark is genuine or forged. In one embodiment, the processor separates the artifacts caused by the damage from the signature signal of the candidate mark from the signature signal artifacts generated by the source marking device. In addition, the processor extracts signature signal data that forms the basis from the candidate mark.

[0044] The term "artifact" of a mark, as used herein, is a feature of a mark generated (controllable or uncontrollable) by the machine or process that created the mark.

[0045] The term "processor" as used herein means a circuit (type of electronic hardware) designed to perform a complex function defined in terms of mathematical logic. Examples of logic circuits include microprocessors, controllers, or application specific integrated circuits. It will be understood that when this disclosure refers to a computer system that performs operations, this also means that a processor integrated with a computing device actually performs the operations.

[0046] Referring now to the drawings, FIG. 1 illustrates an example of an original printed mark ("original mark") 100 to which the various methods described herein can be applied. In this example, the original mark 100 is a 2D barcode. This bar code is a data carrier of information, where the information is encoded as a pattern of bright areas 102 and dark areas 104 of the original mark 100. Possible implementations of 2D barcodes include a rectangular grid, where each cell or “module” (eg, modules 102 and 104) in the grid is either black or white and stores bits of data Represent.

FIG. 2 provides an expanded view of some of the variations that exist in the original mark 100. FIG. 2 shows only the edge 200 between the bright and dark areas of the mark 100. Characteristics such as edge linearity, region discontinuity, and feature shape in the mark 100 are readily apparent. A number of irregularities along the edges of the printed features of the mark are clearly visible. It should be noted that this description is provided for clarity and is not necessarily a required processing step. In some embodiments, edge linearity is a major feature and is therefore used. In some embodiments, characteristics other than edge linearity are used.

FIG. 3 shows an example of a candidate print mark 300 (“candidate mark”). Candidate print marks may be genuine or counterfeit (ie, represent a counterfeit example of the original mark 100 shown in FIG. 1 or a second unique example of the original mark 100 for identification purposes). The candidate mark 300 is a 2D barcode that represents the same decoded information as the original mark 100 of FIG. 1 when read by a 2D barcode reader, even in the case of forgery. When the candidate mark 300 of Fig. 3 is obtained, in one embodiment, large features are identified, and the characteristics of these features are measured and captured as signature data that uniquely identifies the candidate mark 300. This signature data is derived from the mark geometry and the physical and optical properties of the appearance, as in the case of Data that is encoded in the mark), in one embodiment, the characteristics of the original mark that are evaluated to create the signature data evaluate the candidate mark. Are the same properties used in doing so that the two signatures can be directly compared.

FIG. 4 is an enhanced diagram of some of the variations that exist in the candidate mark 300 shown in FIG. FIG. 4 shows only the edge 400 of the mark shown in FIG. 3, which is similar to FIG. Corresponding characteristics such as edge linearity, region discontinuity and feature shape in the mark shown in FIG. 3 are readily apparent.

Proceeding to FIG. 5, some examples of features that can be used to generate signature data for a mark are shown (features 500, 502, 504, and 506). The characteristics of these features that can be measured include average module color or marking strength (eg, feature 500), module position bias (eg, feature 502) with respect to the best fit grid, extraneous printed spots on the mark or not printed It includes the presence or location of a void (eg, feature 504) and the shape of a long continuous edge (linearity) (eg, feature 506). In one embodiment, the measurements performed on these features serve as a primary metric that forms the unique signature of the mark.

[0051] In one embodiment, the metrics obtained from a mark can be weighted as part of forming the mark's signature. For example, the ease with which each of the four metrics shown in FIG. 5 can be extracted depends on the imaging resolution, and the metric is required to extract useful data for each of the four metrics. They can be arranged in the order of resolution. In order from lowest resolution to highest resolution, these are (in one embodiment) module coloring, module position bias, void / mark location and edge shape projection. In other words,
Low resolution -----------------------------> High-resolution module coloring-> module position bias-> void / mark location-> edge Shape projection.

[0052] Increasing image fidelity and resolution allows for increasingly precise analysis using progressively more accurate analysis. For example, in low-resolution images, perhaps only module average coloration and module position bias can be extracted with great confidence, so these results are the original marks (represented by stored real data) More weight is given when determining the signature match of the candidate mark for. With a high resolution image, processing can continue until a fine edge projection metric, and the processor can use this as the highest weight consideration in determining signature matches. If there are discrepancies with the expected signature between the scales), these may be due to mark damage or image capture device artifacts. On the other hand, original mark damage, alteration, or imager artifacts are typically less likely to result in a fake mark being altered to coincide with the edge projection signature metric of the original mark by chance. Thus, edge projections can replace low resolution metrics that support high match confidence if they are highly correlated in dynamic range and exhibiting the right size.

[0053] Further, in one embodiment, the signature metric data is used with the use of error correction information as provided by the standard decoding algorithm of that semiotics (as used in 2D data matrix codes). Is further weighted appropriately. If the data area in the symbol is corrupted by damage to the mark and the area exhibits a discrepancy with the stored signature data, while the other non-corrupted areas match well, The voting weight can be reduced. This mechanism prevents detectable symbol corruption from presenting false negative results in candidate symbol metric comparisons against genuine symbol signature data. The ISO 16022 “Data Matrix Symbol” specification describes how an example of an error correction code (“ECC”) can be distributed within a 2D data matrix, as well as corrupted and uncorrupted regions within the data matrix. Describes how it can be identified.

[0054] Since marking device types are known in advance to exhibit superior or inferior signature characteristics in different attributes for use in generating metric signature data, the marking device type is used to determine the weight profile. Can be pre-weighted. For example, when a genuine mark is created using a thermal transfer printer, the edge projection parallel to the direction of movement of the substrate material has enough signature size to be encoded as part of the genuine signature data. It is known that the possibility is low. This knowledge of the behavior of various marking devices can be used during capture of the original authentic signature data. When used, the metric used when creating a genuine mark signature is weighted, as appropriate, for the known behavior of that particular marking device type, and the resulting enhancement / emphasis suppression mapping is a metric weight. Become a profile.

[0055] FIG. 6 shows a close comparison of the features in the upper left corner of FIGS. As can be seen most clearly in FIG. 6, the two printed marks 100 and 300 of FIGS. 1 and 3 were the case where they were identical in terms of their over-encoded data. However, it includes a large number of differences on a finer scale resulting from a deficiency in the printing process used to apply the mark. These differences are sustainable, typically almost as sustainable as the mark itself, and are practically unique, especially when the many differences that can be found between the symbols of FIGS. 1 and 3 are combined. It is. Furthermore, since the original symbol must be imaged and reprinted with a much higher resolution than the original print, the differences can be difficult to counterfeit while not creating new identifiable printing defects. Although only the upper left corner of the mark is shown here, the distinguishable features between the two marks shown in FIGS. 1 and 3 exist throughout the mark and can be used according to various embodiments. it can.

[0056] Referring to FIG. 7, one embodiment of a computer system 700 includes a central processing unit or processor 702, an output device 704 including an image acquisition device 708, an input device 706, and a random access memory ("RAM") 709. A read only memory (“ROM”) 712 and a magnetic disk or other long term storage (“storage device”) 713 for programs and data. The computer system 700 can have a printer 705 for generating the original mark, or the printer 705 can be a separate device. Computer system 700 can be connected to an external network 708 or other communication medium through an input-output (“I / O”) interface 714 and can be connected to a server 710 having a long-term storage device 716 through network 708. . Although not shown for reasons of simplicity, several similar computer systems can be connected to the server 710.

[0057] Referring to FIG. 8, one embodiment of the present disclosure includes a first computer system 800 that performs capture and storage operations on an original (ie, known to be authentic) mark 802, and a candidate And a second computer system 804 that performs capture, comparison and verification on the mark 806 (ie, it may be genuine or counterfeit). The first computer system 800 includes an image capture device 808 and a signature extraction and encoding processor 810. The image acquisition device 808 captures the image of the original mark 802, generates data relating to the image, and provides the image data to the signature extraction and encoding processor 810. The image extraction and encoding processor 810 can be software running on a processor (such as the processor 702 of the computer system 700) or can be a dedicated processor. The signature extraction and encoding processor 810 stores the signature data on a network accessible mark signature data storage device 812 that can be the long-term storage device 716 of the server 710. In one embodiment, the metrics that make up the signature data are stored as a list sorted in descending order of magnitude and include information identifying their position in the mark from which they were extracted. A network accessible mark signature lookup engine 814, which may be software executing on a processor (such as processor 702 of computer system 700) or a dedicated processor, includes a signature extraction and encoding processor 810 and / or Alternatively, signature data is received from the signature data storage 812.

[0058] The second computer system 804 comprises an image acquisition device 816, a signature extraction and encoding processor 818, and a signature comparison processor 820. Image acquisition device 816 captures the image of candidate mark 806, generates data relating to the image, and provides the image data to signature extraction and encoding processor 818. The signature comparison processor 818 compares the signature extracted by the signature extraction and encoding processor 818 from the (most recently scanned) candidate mark 806 with a signature previously stored in the signature data storage 812 and associated with the original mark 802. can do. In one embodiment, as symbolically indicated by the separation between the upper part of FIG. 8 (for original mark signature capture and storage) and the lower part of FIG. 8 (for candidate mark signature capture, comparison and verification). In addition, the computer system that scans the candidate mark 806 may be different from the computer system that scanned the original mark 802. If they are different, they can either share with access to the signature data storage 812, or a copy of the stored signature data can be retrieved from the original mark capture and signature data storage 812 on the storage system 800. Can be passed to the comparison and verification system 804. In other embodiments, the first computer system 800 and the second computer system 804 are the same system (eg, computer system 700 of FIG. 7).

[0059] Returning to FIG. 7, in one embodiment, the processor 702 may use an image to measure various characteristics (eg, physical characteristics and optical characteristics) of a mark (eg, original mark 100 or candidate mark 300). The image captured by the acquisition device 708 is analyzed, resulting in a set of metrics that contain data regarding mark artifacts. As will be further described, the set of metrics may be one of several sets of metrics that the processor 702 generates for the marks. The processor 702 can take measurements at various locations on the mark. In doing this, the processor 702 can divide the mark into multiple sub-areas (eg, according to industry standards). In one embodiment, if the mark is a 2D barcode, the processor 702 performs measurements on all or a subset of the total number of cells in the mark. Examples of mark characteristics that the processor 702 can measure are: (a) feature shape, (b) feature aspect ratio, (c) feature location, (d) feature size, (e) feature contrast, (f) edge Linearity, (g) region discontinuity, (h) extraneous marks, (i) printing defects, (j) color (eg, brightness, hue or both), (k) coloring, and (l) contrast change. Including. In some embodiments, the processor 702 performs measurements at the same location on a mark-by-mark basis at different locations for different properties. For example, the processor 702 measures the average coloration at a first set of mark locations and at the same first set of locations for subsequent marks, but the second of the locations at that mark and subsequent marks. In some cases, edge linearity may be measured. Two sets of locations (for different characteristics) may be said to be “different” if there is at least one location that is not common to both sets.

[0060] If the mark is a data carrying symbol such as a 2D barcode, various embodiments may utilize additional information embodied by or encoded within the mark. The information to be encoded, for example, a unique or non-unique serial number itself, can then be included as part of the signature data or used to index the signature data for easier retrieval. Further, for 2D barcodes or other data carriers that can establish a quality measure, the processor 702 can extract the information that represents the quality of the mark and is included as part of the signature data.

[0061] Using the quality information, it is possible to detect a change to a genuine candidate mark that may erroneously determine that the mark is counterfeit. This is because these changes may change the signature data of the mark. Some of the quality measurements that can be used include, but are not limited to, unused error correction and fixed pattern damage, as specified in ISO spec 15415 “Data Matrix Rating Process” or other comparable standards. is there. These measures make it possible to detect areas that contribute to the signature data modified by damage to the mark, so that when comparing the signature data of the mark against the stored signature data of the genuine mark, this measure Allows the area to be removed from consideration.

[0062] Referring to FIG. 9, a process performed to capture and store signature data for an original mark (eg, by computer system 700 or 800, or its processor), according to one embodiment, is described. . In step 902, an original mark (eg, a 2D barcode similar to that shown in FIG. 1) is affixed to the object or affixed to a label that is then affixed to the object (eg, by printer 705). It is done.

[0063] In step 904, a mark is acquired by an appropriate imaging or other data acquisition device, such as image acquisition device 808. The imaging device that obtains the mark may take any convenient form such as a camera, machine vision device or scanner. The imaging device can be a conventional device or a device that will be developed in the future. In this embodiment, the imaging device collects data regarding the characteristics of the mark at a level of detail that is significantly finer than the controllable output of the device to which the mark has been applied. In the example shown in FIGS. 1-4, the characteristics include the shape of the boundary between the bright and dark areas with a resolution much finer than the size of the printed 2D barcode module. Other examples of suitable characteristics are described below.

[0064] In step 906, the unique identifier ("UID") included in the unambiguous data of the original mark is decrypted. In one embodiment, if the printer 705 is on the same computer system as the image acquisition device 808, the UID can be passed from one to the other, and there is a need to decrypt the UID from the image acquired by the image acquisition device 808. Avoided. If the original mark does not contain a UID, some other information that uniquely identifies a particular example of the mark can be used in this step.

[0065] In steps 908 and 910, the original mark image is analyzed by the signature extraction and encoding processor 810. For example, the signature extraction and encoding processor 810 extracts quality measurements from the original mark image at step 908 and extracts features from the original mark image at step 910. In step 912, data related to the extracted feature characteristics (eg, metric) is encoded into numeric data and stored in the signature data storage 812 as “signature” data that uniquely identifies the original mark. In one embodiment, the records for each mark are indexed under a unique identifier content (usually a serial number) included in the data that is explicitly encoded in the mark. The records may be stored on a data storage server or device accessible by the network (such as storage device 716) or may be stored locally (such as on storage device 713) if necessary. Copies may be distributed to local storage at multiple locations.

[0066] In steps 914 and 916, candidate signature features are evaluated to ensure that they possess the appropriate size to operate as part of each signature metric. These steps ensure that the features that make up each signature metric possess an actual “signal” for encoding as a distinguishing characteristic of the mark.

[0067] In one embodiment, using the 2D data matrix code as an example, in steps 910, 912, and 914, the four characteristics of the features of the original mark are extracted and sorted by size. As explained above, the image of the mark is acquired so that the features can be processed in electronic form, usually as a color or grayscale image. As a preliminary step, the 2D data matrix is first analyzed as a whole to determine a “best fit” grid that defines the “ideal” location of the boundaries between the cells of the matrix. The candidate feature is then selected by finding the feature that deviates most from the “normal” or “optimal” state of the mark attribute for the particular metric being analyzed. Considering the 2D data matrix code example shown in FIG. 5, some suitable features are:

[0068] The portion of the mark closest to the global average threshold (ie, “lightest” dark module and “darkest”) that distinguishes dark modules from light modules as the average color, coloration or mark intensity is determined by the data matrix readout algorithm "Bright module" (eg feature 500).

[0069] 2. A module (eg, feature 502) that is marked at a position that deviates most from an idealized location as defined by a best-fit grid applied to the mark. Two possible ways to identify these modules are: (a) extract candidate mark module edge positions and their expected positions as defined by the best-fit grid idealized for the mark. (B) extract a histogram of the boundary area between two adjacent modules of opposite polarity (light / dark or dark / light), the sample area is the same for each module relative to the best fit grid Overlap by percentage and evaluate histogram deviation from 50/50 bimodal distribution.

[0070] 3. A foreign mark or void in a module is defined as a module (eg, feature 504) that has a wide range of brightness or pigment density, whether bright or dark. In other words, they are defined as modules that possess coloring levels on both sides of the global average threshold that distinguishes dark modules from light modules, and the best signature candidate is the largest among the outermost dominant modes. A candidate having a bimodal luminance histogram with distance.

[0071] 4. The shape of long continuous edges in a symbol, such as continuity / linearity or discontinuity / nonlinearity (eg, feature 506). One way to measure this attribute and extract this data is by performing a pixel-value luminance value projection, where the projection length of one module offset from the best-fit grid by one-half module is , Extending perpendicular to the grid lines that delimit the edges of the best fit grid for the symbol.

[0072] The 2D data matrix is a good example because it includes square black cells and white cells in which the features described above are easily seen. On the other hand, the same principle can of course be applied to other forms of data encoded or non-data encoded visible marks.

[0073] Once candidate features that match the criteria described above are identified, the candidate features are sorted into a list in order of magnitude in step 914, and then in step 916, the contributing factor for that metric. Is subjected to a size restriction filter by finding the first feature in each list that does not satisfy the minimum size established to be eligible. The threshold can be set to any convenient level that is low enough to include a reasonable number of features that cannot be easily reproduced and high enough to exclude features that are not reasonably durable, or Located near the noise floor of the image acquisition device. In this embodiment, the low size ends of the sorted list are truncated from that point, and the remaining (maximum size) features are along their location within the mark for that metric. Is stored as signature data. Preferably, all features that exceed the truncation threshold are stored, which implicitly includes in the signature information that there are no signature features that exceed the magnitude filter threshold elsewhere in the mark.

[0074] In step 918, based on the type of marking device used to create the original mark, a metric weight profile is stored as part of the signature data.

[0075] In step 920, signature metrics are stored as a sorted list of features in descending order of magnitude. The list entry for each feature includes information for specifying the position in the mark from which the feature is extracted.

[0076] Referring to FIG. 10, a process performed to capture and verify signature data for a candidate mark according to one embodiment (eg, by computer system 700 or 804 or by its processor) is described. . In step 1002, the candidate mark image is acquired by an image acquisition device, such as the image acquisition device 816. In step 1004, explicit data in the candidate mark is decrypted and its UID content is extracted.

[0077] In step 1006, the UID is used to look up previously stored signature metric data for the original mark with that UID. Stored data may be retrieved from a local storage such as storage device 713 or from a long-term storage such as a network accessible data storage server or storage device 716. In the case of a candidate mark that does not include a UID, some other specific information can be obtained in connection with the candidate mark. Alternatively, the entire database of genuine mark signatures (eg, at storage device 713 or storage device 716) is searched after step 1014 below to attempt to locate a genuine signature that matches the candidate mark signature. can do.

[0078] In step 1008, for a 2D barcode or other data carrier for which a quality measurement can be established, the quality measurement 1008 for the candidate mark is similar to that obtained in step 908 for the original mark. Can be obtained. The quality measure can be used in subsequent analysis steps to reduce the weight given to the mark or part of the mark that appears damaged after being applied. Also, if the quality measurement value of the original mark is stored as part of the signature data for the original mark, the stored quality measurement value can be verified against the signature data extracted from the candidate mark.

In step 1010, a large signature feature is extracted from the candidate mark image acquired in step 1002. The entire candidate mark (except for the section that was determined to be ineligible for corruption due to an ECC error) is searched for large features. In addition, the location where the signature data is extracted from the candidate mark can be specified using information identifying the location within the mark from which the original genuine signature data was extracted. This ensures that features that are present in the original mark but not in the candidate mark are noted.

[0080] In step 1012, the signature features are encoded for analysis. In step 1014, the signature data (metric) extracted from the candidate mark is sorted (eg, sorted by size) in the same order as the previously generated list of metrics for the original mark. In step 1016, the candidate signature data is compared with the stored original signature data. There are various ways in which this can be done. In one embodiment, the data is subjected to statistical operations that reveal a numerical correlation between the two data sets. Each metric is subjected to an individual numerical analysis that yields a measure that reflects the individual confidence of the candidate symbol as a genuine item for that metric. If the mark does not contain UID data, alternative specific data is not available and it may be necessary to search a database of similar marks using the procedure discussed below with reference to FIG. is there. For example, in the case of the original mark 100 and the candidate mark 300, it may be necessary to search among all genuine marks having the same obvious pattern of black and white modules. The purpose of the search is to identify or fail to identify a single original mark that is uniquely similar to the candidate mark.

[0081] In step 1018, a metric weight profile is stored as part of the authentic signature data, and this information emphasizes the metric as appropriate for the type of marking device used to create the original authentic mark. And / or used to suppress emphasis.

[0082] In step 1020, if the image acquisition devices used in steps 904 and 1002 have different sensitivities, it may be necessary to adjust the contribution of the signature data to the overall analysis result. For example, the minimum size threshold used for large features may need to be set to an appropriate level for less sensitive image acquisition devices, or certain metrics are generated by the original marking device. It is known that it does not have the proper signature size in the mark, so it may need to be omitted from the analysis suite. In some cases, features recognized in one of the higher resolution categories on the scale shown above may be mistaken for features in different categories by lower resolution scanners. For example, features that are seen as black modules with white voids at high resolution may be seen as “low color modules” at low resolution. Typically, the resolution of the image acquisition device is used in conjunction with a marking device metric weight profile to determine which metric is to be enhanced / suppressed. In this example, the feature is present in the “low pigment” list in the low resolution image, but is present in both the “low pigment” and “empty” lists in the high resolution image.

[0083] If it is desired to explicitly correct for the resolution of the original and / or verification scan, in many cases the resolution will detect a relatively steep drop in the number of artifacts at the scanner's resolution threshold. Can be obtained at the time of verification. Alternatively, if the image acquisition device used to capture the image of the original mark has a lower resolution than the image acquisition device used to capture the image of the candidate mark, derive the resolution or resolution of the scan Other information that can be done can be included as metadata along with the stored signature, similar to the metric weight profile discussed above.

[0084] In step 1022, due to the exclusion, all locations in the mark that are not represented in the sorted list of feature locations that meet the minimum size threshold are missing large signature features when parsing the genuine mark. Expected. This condition checks the signature feature size at all locations within the candidate mark where a partial threshold feature is expected, and if a feature is found that exceeds the threshold minimum value, the result for the appropriate metric goes negative. It is evaluated by adjusting. When large features are found in an area deemed damaged when evaluated for symbol error correction or other quality attributes, depending on the location of the damage relative to the feature extraction point and the nature of the particular metric involved The adjustment is reduced or not performed at all. For example, if a signature feature difference for an original mark is extracted from a module of a candidate mark that is in the vicinity of the damaged module but is not the same as that module, a negative adjustment to the metric attributed to that feature can be Can be reduced by a ratio that reflects the reduced confidence in. This is due to the fact that the module in the vicinity of the known damaged area has suffered damage that affects the metric but falls below the detectable threshold of the symbolic quality or ECC evaluation mechanism. If the difference is extracted directly from the damaged module, or if the metric is one of the types spanning multiple modules and the range includes damaged modules, no adjustment is applied.

[0085] In step 1024, these individual confidence values are then used to determine that the overall confidence in the candidate mark is genuine (or counterfeit). Here, the individual reliability values are appropriately weighted as described above using image fidelity, resolution and symbol damage information.

[0086] In step 1026, it is determined whether the result is clear enough to be acceptable. If the signature data comparison yields intermediate results (eg, individual metrics have inconsistent information that cannot be resolved through the use of a data weighting mechanism) and does not exceed the retry limit (step 1028), The user submitting the symbol for verification is prompted to resubmit another image of the symbol for processing, and the process returns to step 1002. Otherwise, the process ends (process 1030).

[0087] Upon successful completion of the analysis, the results of the comparative analysis are reported in step 1030. The report may pass / fail and can indicate the confidence level in the results. These results can be displayed locally or transferred to a networked computer system or other device for further operation.

Local reference measurement for metric data for environmental tolerance
[0088] In order to further perform robust and accurate signature data extraction in one embodiment, various methods described herein may use area-local criteria within parsed symbols to construct signature data. (Area-local referencing) can be used. This allows for the above-mentioned substrate distortion, non-uniform illumination of candidate symbols when acquired to process non-ideal or low-quality optics, or many other environmental or systematic variables in the acquisition device Provides greater resistance to anything. In one embodiment, metric criteria localization is as follows:

[0089] The average module color, coloring, or mark intensity refers to the nearest neighbor of the opposite module state (dark vs. light, or light vs. dark). If a cell is identified as a large feature with a deviating average color density, the nearest cell may need to be reevaluated by discounting the specific deviating cell as a reference.

[0090] 2. The module grid position bias is referenced to the overall symbol best fit grid and thus has a negative adaptive reference position specification.

[0091] 3. Analysis of extraneous marks or voids in the symbol module uses local color, coloration or mark intensity criteria for the module. In other words, the image luminance histogram in the analyzed module itself provides a reference value for the method applied.

[0092] 4. The projection method used to extract the shape of long continuous edges within a symbol is inherently singular and has negative immunity to variables that have normal effects.

[0093] FIG. 11 illustrates a 1D linear barcode 1100 having features that can be characterized to produce a set of metrics for use as part of the signature data. These features include changes in bar 1102 width and / or spacing between bars 1102, changes in average color, coloration or intensity 1104, voids in black bars 1106 (or black spots in white stripes), or edges of bars 1108 Including irregularities in the shape of

Analysis by autocorrelation method
[0094] In the embodiment described above, the raw list of data for each metric is first subjected to a normalized correlation to the metric set that is matched to the array and index and extracted from the candidate symbols in a similar order. be able to. These correlation results are then used to arrive at a match / non-match decision (genuine vs. counterfeit). To do this, the signature storage includes the original genuine symbol module sort order and the trained metric value itself completed for each metric. In addition to the need for comprehensive storage, raw data is not “normalized”. This is because each metric has its own scale and in some cases unlimited, which complicates the choice of storage bit depth. A typical implementation of the embodiment described above has a stored signature size of about 2 kilobytes.

[0095] Referring now to FIGS. 12-16, an alternative embodiment of the metric post-processing, storage and comparison method is used to extract the original artifact metrics and associate the index-array associated list (modules in symbols). Applied after it becomes available (can be related by location). Based on autocorrelation, the application of this new post-processing method offers several significant advantages compared to the signature of the previous embodiment in at least some situations. The biggest advantage is that the data package size is reduced. For example, a 75% reduction in stored signature data has been realized. More (up to 90% reduction) is possible by applying some smaller additional data compression methods. This dramatic reduction results from the use of autocorrelation, list sorting, and the resulting normalization and data modeling opportunities that allow these mechanisms to be applied to the original artifact data.

Comparison of metrics for original and candidate mark signatures
[0096] Various methods are now described in which signatures for original marks and candidate mark signatures can be compared.

[0097] In the embodiment described above, the analysis of the particular set of metric data includes sorting the raw metrics extracted from the candidate symbols with the raw metrics in a similar order extracted from the real symbols. When taking the form of comparison, the autocorrelation method compares the autocorrelation sequence of the sorted candidate symbol metric data with the (stored) sorted autocorrelation sequence of the genuine symbol data, where Correlate autocorrelation. In one embodiment, the following normalized correlation equation is used:

ここで、ｒは相関結果であり、ｎはメトリックデータリストの長さであり、ｘおよびｙは、本物のメトリックデータセットおよび候補メトリックデータセットである。動作が自己相関として実施されるとき、双方のデータセットｘおよびｙは同じである。

Where r is the correlation result, n is the length of the metric data list, and x and y are the real metric data set and the candidate metric data set. When the operation is performed as autocorrelation, both data sets x and y are the same.

[0098] To generate an autocorrelation sequence, correlation is performed multiple times each time the sequence x is offset by one additional index position relative to the sequence y (recall that y is a copy of x). Wanna) As the offset progresses, the data set must “roll” back to the beginning when the last index in the y data series is exceeded due to the x index offset. This is often most practically accomplished by doubling the y data, “sliding” the x data from offset 0 to offset n, and generating an autocorrelation sequence.

[0099] When performing the autocorrelation approach, the first advantage observed is that the signature data value itself need not be stored as part of the stored data. In autocorrelation, the data series is simply correlated with itself. Thus, where it was previously necessary to deliver both the extraction (sort) order and the authentic signature data value to the verification device for validation, it is now possible to sort / Only the extraction order need be provided.

[0100] Genuine autocorrelation signatures that need to be compared with the candidate symbol results do not need to store genuine data or pass to a verifier. Since the operation of generating a signature is always performed on the stored metric data, the autocorrelation sequence for the original artifact information is always a simple polynomial curve. Thus, instead of having to store the entire autocorrelation sequence of each genuine symbol metric, for each metric, the best fit curve that matches the shape of the genuine autocorrelation result is described (with a predetermined order and accuracy). It is sufficient to store a set of polynomial coefficients.

[0101] In one embodiment, r _xy is calculated. Here, each term x _i is an artifact represented by its size and location, and each term y _i = X _{(i + j)} , where j is 2 from j = 0 to (n−1). Is the offset of one dataset. Since x _i is sorted by magnitude, and the magnitude is the most significant digit of x _i , there is a very strong correlation at or near j = 0 that falls rapidly towards j = n / 2 To do. Since y is a copy of x, j and n−j are interchangeable. Therefore, the autocorrelation sequence always forms a U-shaped curve shown in FIG. This U-shaped curve needs to be symmetric about j = 0 and j = n / 2. Therefore, only half of the curve needs to be calculated. However, in FIG. 13, for the sake of clarity, the entire curve from j = 0 to j = n is shown.

[0102] In practice, it has been found that a 6th order equation using 6 byte floating point values for the coefficients always matches the real data within 1% curve fitting error or "recognition fidelity". That is, if candidate validation is performed using the actual autocorrelation number, and then revalidation is performed on the same mark using a polynomial modeled curve, the resulting match scores are Less than 1%. This is true for both high match scores for genuine candidate marks and low match scores for counterfeit candidate marks. This allows a complete autocorrelation sequence to be represented using only 7 numbers. Assuming that 100 data points are obtained for each metric and there are 6 metrics (known to be a reasonable actual number), this reduces the 600 data values to only 42, There is no loss of symbol discrimination or analysis fidelity. If the individual number is larger, for example, 600 raw numbers are 4-byte integers and 42 polynomial coefficients are 6-byte floating point numbers, almost 90% of the data There is a reduction. In one experimental prototype, 600 single-byte values become 42 4-byte floating point numbers, reducing 600 bytes to 168 bytes, a 72% reduction.

[0103] Next, the stored signature data is explicitly demarcated and normalized. The polynomial coefficients are represented to a fixed accuracy, the autocorrelation data itself is always between -1 and +1 by definition, and the sort order list is simply the module array index location within the parsed symbol. In the case of a 2D data matrix, the module array index is a raster-ordered index of module positions within a symbol and is ordered from the traditional origin datum for its symbology, and thus, according to the definition of matrix symbology Has a defined maximum size. In one common type of 2D data matrix, the starting point is the point where two solid bars that define the lower left border of the grid meet. A standard sorted list length of 100 data points is also established for each metric, giving a predictable, stable and compact signature.

[0104] In one embodiment, the comparison of the genuine signature against the candidate now begins by "reconstructing" the genuine symbol autocorrelation signature by using the stored polynomial coefficients. Next, raw metric data is extracted from the candidate symbols and sorted in the same sort order. This can be shown as part of the genuine signature data if it is not predetermined.

[0105] Next, the candidate metric data is autocorrelated. The resulting autocorrelation sequence can then be correlated to a real autocorrelation curve reconstructed for that metric, or alternatively by calculating the curve fitting error between pairs. Two curves can be compared. This correlation is shown schematically in FIGS. 13 and 16. This final correlation score then becomes the individual “match” score for that particular metric. When complete for all metrics, a “match” score is used to make a genuine / counterfeit decision for the candidate symbol.

[0106] Further, autocorrelation curves can be further used by applying power series analysis to the data by discrete Fourier transform ("DFT").

ここで、Ｘ_ｋはｋ番目の周波数成分である。Ｎはメトリックデータリストの長さであり、ｘはメトリックデータセットである。

Here, _Xk is the kth frequency component. N is the length of the metric data list and x is the metric data set.

Next, the power series of the DFT data is calculated. Next, each frequency component represented by a complex number in the DFT sequence is analyzed for magnitude, and the phase component is discarded. The resulting data describes the distribution of metric data spectral energy from low to high frequencies, which is the basis for further analysis. Examples of these power series are shown schematically in FIGS. 14, 15 and 17.

[0108] Two frequency domain analyzes are used, a measure of energy distribution around the central band frequency of the total spectrum, called kurtosis and distribution bias. Kurtosis is a common statistical operation used to measure the “sharpness” of a distribution, where it signals the presence of closely grouped frequencies with limited bandwidth spread in power series data Useful to do. In one embodiment, the modified kurtosis function can be used as follows.

ここで、

here,

は、べき級数大きさデータの平均であり、ｓは大きさの標準偏差であり、Ｎは解析された離散スペクトル周波数の数である。

Is the average of the power series magnitude data, s is the standard deviation of magnitude, and N is the number of discrete spectral frequencies analyzed.

[0109] The dispersion bias is

として計算され、ここで、Ｎは解析される離散スペクトル周波数の数である。

Where N is the number of discrete spectral frequencies to be analyzed.

[0110] A smooth polynomial curve of real symbol metric signatures (resulting from a sort by magnitude) results in a recognizable property in the spectrum signature when analyzed in the frequency domain. Candidate marks exhibit a similar spectral energy distribution if the mark is genuine when the metric data is extracted in the same order as defined by genuine signature data. In other words, the genuine sort order “matches” the size of the candidate metric. Discrepancies in the sorted sizes, or other overlapping signals (such as duplicate artifacts) tend to appear as high frequency components that do not otherwise exist in the real symbol spectrum, thus increasing symbol reliability. Yield a measure. This addresses the possibility that the counterfeit autocorrelation sequence still meets the minimum statistical match threshold of the genuine mark. This is a small possibility, but the overall range of the data is large compared to the magnitude of the error between the individual data points, and the natural sort order of the dominant metric size is the size of the real symbol. If it happens to be close to the sort order, it can happen in some cases when using normalized correlation. Such a DFT power series distribution characteristic of the signal will reveal the poor quality of matching due to the high frequencies present in the small amplitude matching error of the candidate sequence. Such a condition can indicate a copy of a real symbol. In certain terms, we now expect high kurtosis and high distribution ratio in the spectrum of real symbols.

[0111] The power series distribution information can be used as a measure of "reliability" in the verification of candidate symbols together with the autocorrelation match score.

[0112] FIG. 12 shows a comparison of autocorrelation sequences for a single metric between a genuine item (polynomial approximation) and a candidate symbol (in this case authentic). Here, the correlation between the two autocorrelation sequences exceeds 93%.

[0113] FIG. 13 is a power series from the original genuine autocorrelation data used for FIG. The spectrum can clearly be seen that low frequencies are dominant.

FIG. 14 is a power series similar to FIG. 13 from an image of a real mark obtained by a mobile phone. Although there is some image noise, the overall power spectrum closely matches the real spectral spectrum and has the same dominant low frequency component.

[0115] FIG. 15 shows a comparison of autocorrelation sequences for a single metric between polynomial approximations for genuine (original) marks and candidate marks (here counterfeit). There is considerable discrepancy and the candidate autocorrelation is significantly more jagged than in FIG. The numerical correlation between the two series is low (<5%) and the jagged shape of the data is also seen in the DFT analysis.

FIG. 16 shows the power series obtained from the image of the forged symbol in the plot of FIG. 15 obtained by the mobile phone. Note how the low frequency components decrease with the total spectral energy spread to include a larger portion of the higher frequency range.

Exclude metrics that indicate damage
[0117] In various embodiments, the processor or computer system (such as one or more of those shown in FIGS. 7 and 8) may cause the mark to be damaged before determining whether the mark is genuine. Identify metrics that are likely to represent parts and exclude such metrics. Referring to FIG. 17, there are shown marks used to help illustrate such a process. The exemplary mark in FIG. 17 is a 2D barcode that includes several features. The mark has a left edge finder bar and a bottom edge finder bar. These are solid black bars along the left and bottom edges that are used by the detection equipment to establish the spatial parameters needed to read the information encoded by the mark. The exemplary mark encodes useful information in a pattern of bits delimited by the finder bar, black marks representing binary digits, and white marks. Useful information can include redundant information to correct errors caused by machine damage after the mark is created. Such redundant information can include additional bits that are determined using any suitable error correction code, such as a Reed-Solomon code, a Hamming code, or the like. The use of an error correction code corrects any bit damaged to the extent that it is erroneously read, and thus compares the corrected version of the decoded information with the corrected version of the decoded information, either directly or By doing so, it becomes possible to specify. The marks also include changes such as pinhole spots in the black areas, wobbles at the edges of the black areas, including artifacts of normal operation of the marking equipment that occur randomly and repeatedly in different marks made by the marking equipment.

[0118] FIGS. 18 and 19 show the results of various types of damage, such as dents and scratches that change the black areas of the mark of FIG. As further described in connection with FIGS. 21-25, the signature signal restored from the mark of FIG. 18 differs from that restored from the marks of FIGS. 19 and 20 due to the effects of existing damage.

[0119] In an example where the damage to the candidate mark only makes the mark partially readable, or makes it impossible to read and / or decode data retention symbols etc., only a part of the mark has been damaged. The missing identification feature may be sufficient to identify the mark. When the candidate mark is thus matched to the original mark, the signature of the original mark can be removed from storage and any information embedded in the signature, such as the serial number of the marked item, is damaged. It is possible to restore from the extracted signature, instead of restoring directly from the mark. For this reason, the signature data can be used to uniquely identify the item to which the candidate mark is applied in combination with or without combination with the partially restored encoded symbol information.

[0120] In one embodiment, a processor or computer system (such as one or more of those shown in FIGS. 8 and 9) performs the following procedure. (1) perform a statistical correlation process on the candidate signature versus the original signature; and (2) based on the correlation process, the processor or computer system determines that the candidate signature is not from a genuine mark, (3) Identify the damaged signal that is superimposed on the candidate signature signal, (4) remove the damaged signal, and (5) perform a statistical correlation process on the candidate signature versus the original signature to obtain the remaining signature signal (ie, damaged signal). And (6) if the correlation is sufficiently improved, conclude that there is a match between the original signature and the candidate signature; ) If the correlation does not improve sufficiently, conclude that there is no match between the original signature and the candidate signature.

[0121] To show how much the likelihood of accurate recognition can be improved by removing the damage signal from the signature of the candidate mark, reference is made to Figs. 20-22 illustrate the performance of the correlation technique on the marks of FIGS. 17-19 without removing the damage signal. This is compared with FIGS. 23-25, which illustrate the implementation of the techniques described herein for removing the damage signal. As shown, if there is no damage on the mark (FIG. 17), the results of the two procedures are not significantly different (FIGS. 20 and 23). On the other hand, in the case of damaged marks (FIGS. 18 and 19), the removal of the damage signal results in significant differences (FIGS. 21-24 and FIGS. 22-25).

[0122] Referring to FIG. 26, a process performed by a processor (eg, of computer system 700 or 804) to capture and verify signature data for a candidate mark according to one embodiment is described. At step 2602, the processor receives a captured image of candidate marks from an image acquisition device. In step 2604, the processor uses the image to measure one or more characteristics at multiple locations in the candidate mark, resulting in a first set of metrics, such as the metrics shown in FIG. At step 2606, the processor retrieves from the computer readable memory a second set of metrics representing one or more characteristics measured at a plurality of locations of the original mark. In step 2608, the processor removes a metric indicative of damage to the mark from the first set of metrics, resulting in a trimmed first set of metrics. There are various ways in which the processor can identify metrics that indicate damage to candidate marks. For example, the processor can identify a metric that has a dominant amplitude. In one embodiment, the processor considers metrics that exceed a predetermined threshold to have a dominant amplitude. If the candidate mark includes an error correction redundancy code, the processor can decode the error correction redundancy code to identify one or more damaged regions within the mark. In step 2610, the processor removes a metric corresponding to the metric removed from the first set of metrics from the second set of metrics, resulting in a trimmed second set of metrics.

[0123] For example, in the metric set of FIG. 27, each of the first columns of numbers represents an index value (eg, a raster index), and each of the second columns (eg, associated with an index value). Suppose that it represents a metric value obtained from measuring the characteristic of the mark (at the mark location). Assume further that the “damage noise” threshold is 185. In other words, the processor considers any metric value above 185 to indicate damage to the candidate mark. In this example, the processor removes the metric associated with index number 169 from both the first set of metrics and the second set of metrics.

[0124] At step 2612, the processor compares the trimmed first set of metrics to the trimmed second set of metrics. At step 2614, the processor determines whether the candidate mark can be verified as authentic based on the comparison.

[0125] If, in step 2614, based on the comparison, the processor verifies that the candidate mark is authentic, the processor alerts this to the user in step 2616. On the other hand, if the processor cannot verify that the mark is authentic and the retry limit has not been reached (step 2618), the processor will resubmit another image of the candidate mark for processing. Prompt and the process returns to step 2608. On the other hand, if a retry limit is reached, eg, a predetermined limit on the number of metrics removed from the candidate metric data set, the process ends. In one embodiment, the predetermined limit is 30% or about 30% of the metric data set. For example, in the second iteration, using the trimmed first set of candidate mark metrics, at step 2608, the processor may determine that the metric associated with index number 165 (the metric associated with index number 169 is already present). Removed), generating a trimmed third set of metrics, removing the corresponding metric from the trimmed second set, and generating a trimmed fourth set of metrics. The processor then performs a comparison of the third and fourth sets of metrics, and so forth.

[0126] Referring to FIG. 28, a process performed by a processor (eg, in the computer system 700 or 804) to capture and verify signature data for a candidate mark is described, according to another embodiment. In step 2802, the processor receives an image of the candidate mark from the image acquisition device. In step 2804, the processor uses the image to measure one or more characteristics at multiple locations in the candidate mark, resulting in a first set of metrics. At step 2806, the processor retrieves from computer readable memory a second set of metrics representing one or more characteristics measured at a plurality of locations in the original mark. In step 2808, the processor compares the first set of metrics to the second set of metrics. At step 2810, the processor determines whether the candidate mark can be verified as authentic based on a comparison of the first set of metrics and the second set of metrics. At step 2810, based on the comparison of the first set of metrics and the second set of metrics, the processor determines that the mark can be verified as authentic, and the process then moves to step 2812. In step 2812, the processor notifies the user that the mark is authentic. In step 2810, if the processor determines that the candidate mark cannot be verified as authentic, it executes steps 2814, 2816 and 2818 which are identical to steps 2608, 2610 and 2612 of FIG. The processor also performs steps similar to steps 2614 and 2618 (omitted from FIG. 28 for clarity) to generate a trimmed third set of metrics and a trimmed fourth set of metrics.

[0127] Referring to FIG. 29, a process performed by a processor (eg, in the computer system 700 or 804) to capture and verify signature data for a candidate mark according to another embodiment is described. In step 2902, the processor receives an image of the candidate mark from the image acquisition device. In step 2904, the processor uses the image to measure one or more characteristics at multiple locations in the candidate mark, resulting in a total set of metrics. In step 2906, the processor subdivides the entire set of metrics into multiple sets of metrics including the first set of metrics. In step 2908, the processor retrieves from the computer readable memory a second set of metrics representing one or more characteristics measured at a plurality of locations on the original mark. In step 2910, the processor removes the metric having the dominant amplitude from the first set of metrics, resulting in a trimmed first set of metrics. At step 2912, the processor removes a metric corresponding to the metric removed from the first set of metrics from the second set of metrics, resulting in a second trimmed set of metrics. At step 2914, the processor compares the trimmed first set of metrics to the trimmed second set of metrics. In step 2916, the processor determines whether the mark can be verified as authentic based on the comparison.

[0128] In another embodiment, the processor divides the candidate metric set into subsets by the region of the mark and removes the metric equivalent of the original mark in removing the metric expected to be the result of damage to the candidate mark. Exclude at least one subset that is poorly correlated with the data set. The processor can further repeat the segmentation and removal process, and the candidate mark creates a genuine mark when the remaining subdivision correlation reaches a threshold after excluding at least one subset with poor correlation Judge that it is highly possible that the device was manufactured.

[0129] According to one embodiment, genuine markings that have been captured and reduced to a verification metric data set indexed by amplitude for genuine marking are based on changes in genuine markings and candidate markings. Analyzing a verification metric data set indexed by amplitude for damaged candidate markings, reduced from captured images of candidate markings that are one of genuine markings and counterfeit markings The process performed by the processor to identify by means of removing the verification metric with the dominant amplitude from the verification metric set for candidate marking, and for the candidate marking from the verification metric set for genuine marking Validation metrics Generating a trimmed verification metric set for genuine marking and a trimmed verification metric set for candidate marking by removing the verification metric corresponding to the verification metric having the dominant amplitude in the Comparing the trimmed verification metric set and identifying whether the candidate marking is likely to be produced by the marking device that produced the authentic mark based on the comparison.

[0130] In one variation, the processor-implemented method first determines that the candidate marking is likely not made by the marking device that made the authentic mark before the removing step. In addition. In another variation, the method implemented by the processor is performed on a coded mark that includes error correction redundancy to decode the error correction redundancy in the mark to identify damaged areas in the mark. And removing further includes excluding verification metrics originating from the damaged area. In yet another variation, the processor-implemented method further comprises dividing the verification metric set into subsets by the region of the mark, and removing removing at least one subset that is poorly correlated. In addition. This variation further includes repeating the dividing and removing step, and the candidate mark created a real mark when the remaining sub-correlation reaches a threshold after excluding at least one subset with poor correlation It further includes determining that it is likely that the marking device has been produced.

[0131] According to one embodiment, when the mark is damaged, the damage signal is superimposed on the signature signal, overwhelming the signature signal. The portion of the image of the mark being analyzed is, for example, the linearity (or linearity from a known straight edge) in the mark, such as one of the finder bars in the two-dimensional barcodes in FIGS. Deviation). Thus, when it is determined that the mark is not temporarily originating from a real source, further steps are taken as follows. First, it identifies and removes larger damage signals that are superimposed on the entire signature signal. Next, the signature is evaluated based on the remaining signature signal. If the damaged signal is removed from the signature signal and the correlation improves sufficiently, a match between the original signature and the candidate signature is reported by the method. If the damaged signal is removed and the correlation fails to improve sufficiently and falls, the method confirms that there is no match between the original signal and the candidate signal.

[0132] When an original mark is applied to an original item and / or the original item is added to the original object, the mark or item can include information about the item or object. In that case, the methods and systems described above may include verification information regarding the item or object contained in the mark or item, even if the underlying item or object is not physically exchanged or changed. For example, if an object is marked with an expiration date, it may be desirable to reject the object with the changed expiration date as “not authentic” even if the object itself is the original object. Embodiments of the present system and method generate the result, for example, as a printing defect, if an artifact used for verification is found on the expiration date. Other information such as lot number and other product tracking data can be verified as well.

[0133] Various embodiments have been described in terms of obtaining an entire 2D barcode for signature data. On the other hand, the mark can be divided into smaller zones. If the original mark is large enough and has sufficient artifacts that are potential signature data, then only one or fewer than all zones can be obtained and processed. If more than one zone is acquired and processed, signature data from different zones can be recorded separately. This is particularly useful when the mark is symbol-encoded data that uses error correction, which involves a smaller zone than the entire symbol. Next, if the error correction indicates that some of the candidate symbols have been damaged, the signature data from the damaged portion can be ignored.

[0134] Although the embodiments have been described primarily in terms of distinguishing an original mark (and the original item on which the mark is implicitly attached or attached) from a counterfeit copy of the mark, the method, apparatus and product Can be used for other purposes, including distinguishing between different instances of the original mark (and item).

[0135] Certain embodiments where, for reasons of simplicity, the artifact is a defect in printing of a printed mark that is applied directly to the item being verified or to a label applied to the object being verified Explained. On the other hand, as already mentioned, any feature that is sufficiently detectable and permanent and difficult to replicate can be used.

[0136] Some of the embodiments have been described as using a database of signature data for authentic items. Among them, the search is performed on signature data that at least partially matches the signature data extracted from the candidate mark. On the other hand, if a candidate item is identified as a specific real item in some other way, a search may not be necessary and the signature data extracted from the candidate mark is stored for the specific real item. The signature data can be directly compared.

[0137] With reference to FIG. 30, another computer hardware environment is described in which the various methods disclosed herein may be implemented. An exemplary computer hardware environment can generally be described as a computer network of any suitable size and range that supports and includes the operating elements and structures on which the exemplary methods and apparatus described below are relied upon. Several common operating elements and structural features will now be described in connection with FIG.

[0138] Various computing devices are interconnected for communication through a computer network having any suitable hardware configuration, such as the global Internet computer network 3000. Computing devices include the following devices: mobile devices 3001, portable and fixed computing devices 3002, content, software, software as a service (SaaS), storage and other resource servers 3003, interconnects, switches and routers, etc. Communication resources 3004, as well as other computing resources 3005. Mobile devices, portable fixed computing devices, switches, routers, and servers typically execute software instructions to accomplish a central processing unit (“CPU”), microprocessor, microcontroller, or the intended task used. Includes similar elements. Local instructions and local data are stored in the appropriate form of computer storage and computer memory, including both temporary and non-transitory media and / or signals. The device can include input peripherals, display peripherals, and other peripherals that are integrated into or connected to the device.

[0139] Mobile device 3001 may include a device that integrates a wireless mobile telephone service with a mobile data service connected to the Internet. Examples of such devices include smartphones made by different manufacturers that operate with different operating systems on different carriers. Mobile devices may also include tablets and other devices that are intended to operate from a wide range and a wide variety of locations using cellular, Wi-Fi, and other suitable communication links. it can. Mobile devices can be integrated into wearable formats, eyeglass frames and the like, vehicles, and the like.

[0140] The method described above may be performed using any of the devices implemented by the described processor. Different parts of the method can be performed using different processors. For example, scanning and learning of a genuine mark signature can be performed by a stable computer located where the marking is performed. In the same example, the verification can be performed by a mobile device located where the marked article is consumed, transported, purchased, sold, etc.

Any one of the devices, computers, servers, and the like described above can be a general-purpose computer system 3100 as shown in FIG. Computer system 3100 can include a processor 3103 connected to one or more memory devices 3104, such as a disk drive, memory, or other device for storing data. Memory 3104 is typically used for storing programs and data during operation of computer system 3100. The components of computer system 3100 include one or more buses (eg, between components integrated within the same machine) and / or a network (eg, between components residing on separate discrete machines). Can be coupled by an interconnection mechanism 3104 that can. The interconnection mechanism 3105 allows communications (eg, data, instructions) to be exchanged between system components of the system 3100.

[0142] The computer system 3100 includes one or more input devices 3102 such as a keyboard, mouse, trackball, microphone, touch screen, and one or more output devices 3101 such as printing devices, display screens, speakers. Also equipped. Further, the computer system 3100 can include one or more interfaces (not shown) that connect the computer system 3100 to a communication network (in addition to or instead of the interconnection mechanism 3105). Depending on the specific application to be performed, one or more of the described components can optionally be omitted, or one or more of the described components can be highly specialized and specified For example, a storage system may not have separate input and output devices, which can be a high-speed bus or data to move data and instructions between the storage system and the data consumer For communication systems that use networks It can be combined to have.

[0143] The storage system 3106, shown in more detail in FIG. 32, typically includes a non-volatile recording medium 3201 that can be read and written by a computer, in which signals defining instructions are stored. These instructions together form a program executed by a processor or information stored on or in the medium 3201 processed by the program. The medium can be, for example, a disk or a flash memory. Optionally, the medium can be read-only, so it stores only instructions and static data for performing specialized tasks. In general, in operation, the processor causes data to be read from the non-volatile recording medium 3201 into another memory 3202 that allows faster access to information by the processor than the medium 3201 does. The memory 3202 is typically a volatile random access memory such as dynamic random access memory (“DRAM”) or static memory (“SRAM”). The memory 3202 can be located in the storage system 3106 as shown, or in the memory system 3104, although not shown. The processor 3103 normally manipulates data in the integrated circuit memories 3104 and 3202 and copies the data to the medium 3101 after the processing is completed. A wide variety of mechanisms are known for managing the movement of data between the media 3201 and the integrated circuit memory elements 3104, 3202, and the disclosure is not limited thereto. The present disclosure is not limited to a particular memory system 3104 or storage system 3106.

[0144] A computer system may include specially programmed, dedicated hardware, such as an application specific integrated circuit ("ASIC"). Aspects of the present disclosure can be implemented in software, hardware or firmware, or any combination thereof. Further, such methods, operations, systems, system elements and components thereof may be implemented as part of the computer system described above or as independent components.

[0145] Although the computer system 3100 is illustrated by way of example as one type of computer system that can implement various aspects of the present disclosure, the aspects of the present disclosure are illustrated in FIG. It is to be understood that the invention is not limited to being implemented. Various aspects of the disclosure may be implemented in one or more computers having different architectures or components than those shown in FIG.

[0146] The computer system 3100 can be a general-purpose computer system that can be programmed using a high-level computer programming language. The computer system 3100 can also be implemented using specially programmed general purpose hardware. In computer system 3100, processor 3103 can be any suitable processor for the task at hand. An execution system or operating system in which work programs are hierarchized can control the processor. Any suitable execution system and operating system can be used.

[0147] The processor and operating system together define a computer platform for application programs in which high-level programming languages are written. It should be understood that the present disclosure is not limited to a particular computer system platform, processor, operating system or network. It should also be apparent to those skilled in the art that the present disclosure is not limited to a particular programming language or computer system. In addition, it should be understood that other suitable programming languages and other suitable computer systems may be used.

[0148] One or more portions of a computer system may be distributed across one or more computer systems coupled to a communication network. These computer systems can also be general-purpose computer systems. For example, various aspects of the disclosure may perform overall tasks as part of one or more computer systems (eg, servers) or distributed systems configured to provide service to one or more client computers Can be distributed among one or more computer systems configured to do so. For example, various aspects of the present disclosure can be performed on a client server or multi-tier system that includes components distributed among one or more server systems that perform various functions according to various embodiments of the present disclosure. . These components include executable code, intermediate code (eg, IL) or interpreter code (eg, Java) that communicates over a communication network (eg, the Internet) using a communication protocol (eg, TCP / IP). can do.

[0149] It is to be understood that this disclosure is not limited to execution in any particular system or group of systems. It should also be understood that the present disclosure is not limited to any particular distributed architecture, network or communication protocol.

[0150] Various embodiments of the present disclosure can be programmed using an object oriented programming language such as SmallTalk, Java, C ++, Ada or C # (C-Sharp). Other object-oriented programming languages may also be used. Alternatively, function, script and / or logic programming languages may be used. Various aspects of the present disclosure provide HTML, XML, that render aspects of a graphical user interface (“GUI”) or perform other functions when viewed in a window of a browser program (eg, a browser program window). Or a document generated in another format). Various aspects of the present disclosure can be implemented as programmed or unprogrammed elements, or any combination thereof.

[0151] In view of the many possible embodiments in which the principles of the present discussion can be applied, the embodiments described herein with respect to the drawings are intended to be examples only, and the scope of the claims may be applied. It should be appreciated that it should not be construed as limiting the scope. Accordingly, the techniques described herein contemplate all embodiments that may fall within the scope of the following claims and their equivalents.

Claims (19)

A method for verifying the reliability of a mark,
Receiving an image of a candidate mark from an image acquisition device;
Using the image to measure one or more properties at a plurality of locations in the candidate mark, resulting in a first set of metrics;
Removing from the first set of metrics a metric indicative of damage to the candidate mark, resulting in a trimmed first set of metrics;
Retrieving from the computer readable memory a second set of metrics representing one or more characteristics measured at a plurality of locations in the original mark;
Removing a metric corresponding to the metric removed from the first set of metrics from the second set of metrics, resulting in a trimmed second set of metrics; When,
Comparing the trimmed first set of metrics to the trimmed second set of metrics;
Determining whether the candidate mark can be verified as authentic based on the comparison;
Including a method.   The method of claim 1, wherein removing the metric indicative of damage to the candidate mark comprises removing a metric having a dominant amplitude.   The method of claim 2, further comprising identifying the highest metric in the first set of metrics as having the dominant amplitude.   The method of claim 2, further comprising identifying a metric that is highest in the first set of metrics and exceeds a predetermined threshold as having the dominant amplitude.   The method of claim 2, further comprising identifying the metric having the dominant amplitude based on whether the metric exceeds a predetermined threshold. If it is determined based on the comparison that the mark is not authentic, the method further includes performing a further action, the further action comprising:
Removing from the trimmed first set of metrics a metric indicative of damage to the candidate mark, resulting in a trimmed third set of metrics;
Removing a metric corresponding to the metric removed from the trimmed first set of metrics from the trimmed second set of metrics, resulting in a metric trimmed fourth Get a pair, step,
Comparing the trimmed third set of metrics with the trimmed fourth set of metrics;
Determining whether the mark can be verified as authentic based on a comparison of the trimmed third set of metrics and the trimmed fourth set of metrics;
The method of claim 1 comprising:   The removing, comparing, and determining for a trimmed continuous set of metrics until a metric remaining in the trimmed set of metrics for the candidate mark is less than a predetermined threshold amount. The method of claim 6, further comprising repeating the steps. Further comprising decoding error correction redundancy of the candidate mark to identify a damaged portion of the candidate mark;
The method of claim 1, wherein removing the metric from the first set of metrics comprises removing a metric corresponding to the damaged portion from the first set of metrics. A method for verifying the reliability of a mark,
Receiving an image of a candidate mark from an image acquisition device;
Using the image to measure one or more properties at a plurality of locations in the candidate mark, resulting in a first set of metrics;
Retrieving from the computer readable memory a second set of metrics representing one or more characteristics measured at a plurality of locations in the original mark;
Comparing the first set of metrics with the second set of metrics;
Determining whether the candidate mark is authentic based on the comparison of the first set of metrics and the second set of metrics;
Performing a further step if the candidate mark is determined not to be authentic based on the comparison of the first set of metrics and the second set of metrics, the further steps comprising: Is
Removing from the first set of metrics a metric having a dominant amplitude, resulting in a trimmed first set of metrics;
Removing a metric corresponding to the metric removed from the first set of metrics from the second set of metrics, resulting in a trimmed second set of metrics; and ,
Comparing the trimmed first set of metrics to the trimmed second set of metrics;
Determining whether the mark can be verified as authentic based on the comparison of the trimmed first set of metrics and the trimmed second set of metrics;
Including a method. The method of claim 9 , further comprising identifying a highest metric in the first set of metrics as having the dominant amplitude. The method of claim 9 , further comprising identifying a metric that is highest in the first set of metrics and exceeds a predetermined threshold as having the dominant amplitude. The method of claim 9 , further comprising identifying the metric having the dominant amplitude based on whether the metric exceeds a predetermined threshold. If the mark is determined not to be authentic based on the comparison between the first set of metrics and the second set of metrics, the method further comprises performing further actions The operation is
From the first set that is trimmed of the metric, the method comprising the steps of removing a metric indicating damage to the candidate mark, to obtain a third set that is trimmed resulting metric a first step,
Removing a metric corresponding to the metric removed from the trimmed first set of metrics from the trimmed second set of metrics, resulting in a metric trimmed fourth A second step of obtaining a set;
A third step of comparing the trimmed third set of metrics with the trimmed fourth set of metrics;
Determining whether the mark can be verified as authentic based on the comparison of the trimmed third set of metrics and the trimmed fourth set of metrics ; Steps,
The method of claim 9, comprising: The first step, the second step, for a trimmed continuous set of metrics until the metric remaining in the trimmed set of metrics for the candidate mark is less than a predetermined threshold amount ; The method of claim 13 , further comprising repeating the third step and the fourth step . Further comprising decoding error correction redundancy of the candidate mark to identify a damaged portion of the candidate mark;
The method of claim 1, wherein removing the metric from the first set of metrics comprises removing a metric corresponding to the damaged portion from the first set of metrics. A method for verifying the reliability of a mark,
Receiving an image of a candidate mark from an image acquisition device;
Using the image to measure one or more characteristics at a plurality of locations in the candidate mark, resulting in a total set of metrics;
Subdividing the entire set of metrics into a plurality of sets of metrics including a first set of metrics;
Retrieving from the computer readable memory a second set of metrics representing one or more characteristics measured at a plurality of locations on the authentic mark;
Removing from the first set of metrics a metric having a dominant amplitude, resulting in a trimmed first set of metrics;
Removing a metric corresponding to the metric removed from the first set of metrics from the second set of metrics, resulting in a trimmed second set of metrics; When,
Comparing the trimmed first set of metrics to the trimmed second set of metrics;
Determining whether the mark can be verified as authentic based on the comparison;
Including a method. A computer system comprising a processor for performing an operation, the operation comprising:
Receiving an image of a candidate mark from an image acquisition device;
Using the image to measure one or more properties at a plurality of locations in the candidate mark, resulting in a first set of metrics;
Removing from the first set of metrics a metric indicative of damage to the candidate mark, resulting in a trimmed first set of metrics;
Retrieving from the computer readable memory a second set of metrics representing one or more characteristics measured at a plurality of locations of the original mark;
Removing a metric corresponding to the metric removed from the first set of metrics from the second set of metrics, resulting in a trimmed second set of metrics; When,
Comparing the trimmed first set of metrics to the trimmed second set of metrics;
Determining whether the candidate mark can be verified as authentic based on the comparison;
Including a computer system. The computer system of claim 17 , wherein removing the metric indicative of damage to the candidate mark comprises removing a metric having a dominant amplitude . The processor is
The computer system of claim 18 , further performing an operation comprising identifying the highest metric in the first set of metrics as having the dominant amplitude.

Images